HemaChat: Lessons Learned from Developing a Multi-Agent Healthcare Chatbot for Sickle Cell Disease Care

• David Eduardo Pereira UFCG
• Herman Martins Gomes UFCG
• Claudio E. C. Campelo UFCG
• Eliane Cristina Araújo UFCG

DOI: https://doi.org/10.5753/se4as.2026.32997

Resumo

Software development and applications have changed significantly with the widespread adoption of Artificial Intelligence (AI), particularly Large Language Models (LLMs). Conversational chatbots now enable intuitive interaction through natural language. This industrial experience report presents the development of a healthcare platform centered on an AI-powered chatbot, named Hemachat, built on an LLM-based multi-agent system. This solution aims to support both healthcare professionals and patients with Sickle Cell Disease (SCD) by assisting with medical follow-up, tracking laboratory examinations and consultation schedules, answering general questions about the disease, and supporting patient triage. Furthermore, it generates treatment plans that can assist physicians in the clinical decision-making process. This report focuses on the development steps, the software architecture, and the major software engineering decisions that shaped the system. It further discusses key challenges, including privacy and security, testing AI-powered conversational systems, and ensuring reliable and clinically safe behaviour in a safety-critical healthcare environment. While reporting the development experience and identifying open challenges, we aim to provide practical insights to support developers of multiagent LLM-based healthcare applications.

Palavras-chave: Large Language Model, Multi-Agent System, Sickle Cell Disease

Referências

Amugongo et al. 2025. Retrieval augmented generation for large language models in healthcare: A systematic review. PLOS Digital Health 4 (06 2025), 1–33.

Franklin et al. 2026. AI Agent Traps. SSRN Electronic Journal (8 March 2026).

Leite et al. 2026. LLM-BasedMulti-Agent Systemwith Retrieval-Augmented Generation for Medical Care Planning Generation in Sickle Cell Disease. In Proceedings of the 17th International Conference on Computational Processing of Portuguese (PROPOR 2026) - Vol. 2. Salvador, Brazil.

Mutisya et al. 2025. Mind the Gap: Evaluating the Representativeness of Quantitative Medical Language Reasoning LLM Benchmarks for African Disease Burdens.

Reich et al. 2022. An Integrative Review: The Evolution of Provider Knowledge, Attitudes, Perceptions and Perceived Barriers to Caring for Patients with Sickle Cell Disease 1970–Now. Journal of Pediatric Hematology/Oncology Nursing 40, 1 (July 2022), 43–64.

Roman et al. 2025. ChatChecker: A Framework for Dialogue System Testing and Evaluation Through Non-cooperative User Simulation. arXiv:2507.16792 [cs.AI]

Md. Asraful Haque. 2025. LLMs: A game-changer for software engineers? Bench-Council Transactions on Benchmarks, Standards and Evaluations 5, 1 (2025), 100204.

Li Liu and Vincent G. Duffy. 2023. Exploring the Future Development of Artificial Intelligence (AI) Applications in Chatbots: A Bibliometric Analysis. International Journal of Social Robotics 15, 5 (2023), 703–716.